Protocol v1
How every engagement runs
- Input
- Your brand, your customers' real questions, your site.
- Process
- Observe, hypothesize, change one thing, retest.
- Output
- Fixes that are checkable, and results reported as they are.
About SearchNovaSF
A small GEO and AEO lab with a written protocol
SearchNovaSF helps lawful adult brands (sexual wellness, adult retail, subscription platforms, dating and nightlife) get described accurately in AI answers. We work like a lab: observe, form a hypothesis, change one thing, retest. This page is that protocol, and the rules we do not break.
The protocol, step by step
- 01Observe
Your customers' questions go to each engine, and every answer and source is recorded exactly.
- 02Hypothesize
A weak or wrong answer gets a named likely cause: access, eligibility, wording, evidence or missing facts.
- 03Change one thing
One fix at a time goes onto your site or profiles, logged with what changed and when.
- 04Retest
Over the following weeks the same questions run again, and results are reported honestly, including when nothing moved.
The checks behind the first step are in our readiness checklist, and the services built on it are adult GEO, adult AEO and AI citation optimization.
Lab rules we do not break
- Invent statistics, quotes, reviews or citations
- Hide text or instructions for AI crawlers
- Post disguised content on forums, encyclopedias or review sites
- Promise that any engine will cite or recommend you
- Accept clients who advertise paid sexual services
People rely on AI answers only as long as those answers are trustworthy. Tricks that fool an engine for a while erode that trust and, sooner or later, the brand behind them.
New, and saying so
SearchNovaSF opened in 2026, so this site shows no logos or quotes from clients. Any that appear later will be earned, and shared only with permission. Prices are on the pricing page, and a free readiness audit shows how we work on your own brand.
How we calibrate our own measurements
AI answers are noisy. The same question can produce a different answer an hour later, so a single screenshot proves very little. We ask each question several times in each engine, from a clean session and the buyer's country, and we keep the question list fixed from month to month so that a change in the results means something. When a result moves, we look for the boring explanations first, such as an engine update or a change in our own method, before crediting any fix.
Sources on the bench
We work from primary documents wherever they exist: Google's documentation on AI features, crawling, structured data and its 2026 Search Console changes; the crawler documentation that OpenAI and other AI companies publish; and the GEO paper by Aggarwal and colleagues. Industry studies, such as the Ahrefs and Pew work on citations and clicks, appear with their names, dates and what they measured, including where their methods differ. Every report on our lab shelf carries the date it was last checked.
What the lab cannot see
- How each engine ranks its sources internally. We observe the outputs and test changes; we do not have access to the systems.
- Answers people get when signed in with their own history, which can differ from the clean sessions we test in.
- Anything legal. We summarize rules where they shape pages, and your counsel decides what applies to you.
- Your production code. We write precise tickets; your developers make the changes.
Saying this plainly is part of the protocol. A result we cannot explain is reported as unexplained, not dressed up as a win.
The first month in the lab
Once the free readiness audit is done and you decide to continue, the first week sets up access and agrees the question list. The second records how each engine answers those questions today, run more than once. The third turns the technical findings into tickets, starting with anything that stops engines reaching or reading your pages. The fourth begins the first page rewrites and fixes the brand facts sheet. By day thirty you have a baseline, a ranked list of fixes and the first changes live, and the work continues on the plan set out under pricing.
Questions
Q.01
Why focus on adult brands?
Because AI engines handle adult topics cautiously, and general AI search advice rarely accounts for that. Lawful adult brands need factual, safe-for-work content and careful testing to be described accurately.
Q.02
Is the SF in the name a location?
No, it is simply part of the brand. The team is remote; the legal entity behind the studio is listed in the terms of service.
Q.03
Do you have case studies?
Not yet. We are new, and we only publish results with a client's written agreement. Our lab reports and a free readiness audit are the best way to judge the work.
Q.04
Why do you test each question more than once?
Because AI answers vary from one run to the next. A single answer can be a fluke; the same result across several runs and several months is something worth acting on.
Next experiment
What do AI answers say about your brand today?
- Hypothesis
- AI engines describe you less accurately than your site could support.
- Method
- We ask the questions your customers ask, across the main AI engines.
- Result
- A short readiness report with what to fix first. Free.